netcdf-c/libnczarr/zdim.c
Dennis Heimbigner df3636b959 Mitigate S3 test interference + Unlimited Dimensions in NCZarr
This PR started as an attempt to add unlimited dimensions to NCZarr.
It did that, but this exposed significant problems with test interference.
So this PR is mostly about fixing -- well mitigating anyway -- test
interference.

The problem of test interference is now documented in the document docs/internal.md.
The solutions implemented here are also describe in that document.
The solution is somewhat fragile but multiple cleanup mechanisms
are provided. Note that this feature requires that the
AWS command line utility must be installed.

## Unlimited Dimensions.
The existing NCZarr extensions to Zarr are modified to support unlimited dimensions.
NCzarr extends the Zarr meta-data for the ".zgroup" object to include netcdf-4 model extensions. This information is stored in ".zgroup" as dictionary named "_nczarr_group".
Inside "_nczarr_group", there is a key named "dims" that stores information about netcdf-4 named dimensions. The value of "dims" is a dictionary whose keys are the named dimensions. The value associated with each dimension name has one of two forms
Form 1 is a special case of form 2, and is kept for backward compatibility. Whenever a new file is written, it uses format 1 if possible, otherwise format 2.
* Form 1: An integer representing the size of the dimension, which is used for simple named dimensions.
* Form 2: A dictionary with the following keys and values"
   - "size" with an integer value representing the (current) size of the dimension.
   - "unlimited" with a value of either "1" or "0" to indicate if this dimension is an unlimited dimension.

For Unlimited dimensions, the size is initially zero, and as variables extend the length of that dimension, the size value for the dimension increases.
That dimension size is shared by all arrays referencing that dimension, so if one array extends an unlimited dimension, it is implicitly extended for all other arrays that reference that dimension.
This is the standard semantics for unlimited dimensions.

Adding unlimited dimensions required a number of other changes to the NCZarr code-base. These included the following.
* Did a partial refactor of the slice handling code in zwalk.c to clean it up.
* Added a number of tests for unlimited dimensions derived from the same test in nc_test4.
* Added several NCZarr specific unlimited tests; more are needed.
* Add test of endianness.

## Misc. Other Changes
* Modify libdispatch/ncs3sdk_aws.cpp to optionally support use of the
   AWS Transfer Utility mechanism. This is controlled by the
   ```#define TRANSFER```` command in that file. It defaults to being disabled.
* Parameterize both the standard Unidata S3 bucket (S3TESTBUCKET) and the netcdf-c test data prefix (S3TESTSUBTREE).
* Fixed an obscure memory leak in ncdump.
* Removed some obsolete unit testing code and test cases.
* Uncovered a bug in the netcdf-c handling of big-endian floats and doubles. Have not fixed yet. See tst_h5_endians.c.
* Renamed some nczarr_tests testcases to avoid name conflicts with nc_test4.
* Modify the semantics of zmap\#ncsmap_write to only allow total rewrite of objects.
* Modify the semantics of zodom to properly handle stride > 1.
* Add a truncate operation to the libnczarr zmap code.
2023-09-26 16:56:48 -06:00

276 lines
8.1 KiB
C

/* Copyright 2003-2019, University Corporation for Atmospheric
* Research. See the COPYRIGHT file for copying and redistribution
* conditions. */
/**
* @file @internal This file is part of netcdf-4, a netCDF-like
* interface for NCZ, or a ZARR backend for netCDF, depending on your
* point of view.
*
* This file includes the ZARR code to deal with dimensions.
*
* @author Dennis Heimbigner, Ed Hartnett
*/
#include "zincludes.h"
/**
* @internal Dimensions are defined in attributes attached to the
* appropriate group in the data file.
*
* @param ncid File and group ID.
* @param name Name of the new dimension.
* @param len Length of the new dimension.
* @param idp Pointer that gets the ID of the new dimension. Ignored
* if NULL.
*
* @return ::NC_NOERR No error.
* @return ::NC_EBADID Bad ncid.
* @return ::NC_EMAXNAME Name is too long.
* @return ::NC_EBADNAME Name breaks netCDF name rules.
* @return ::NC_EINVAL Invalid input.
* @return ::NC_EPERM Read-only file.
* @return ::NC_EUNLIMIT Only one unlimited dim for classic model.
* @return ::NC_ENOTINDEFINE Not in define mode.
* @return ::NC_EDIMSIZE Dim length too large.
* @return ::NC_ENAMEINUSE Name already in use in group.
* @return ::NC_ENOMEM Out of memory.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
NCZ_def_dim(int ncid, const char *name, size_t len, int *idp)
{
NC *nc;
NC_GRP_INFO_T *grp;
NC_FILE_INFO_T *h5;
NC_DIM_INFO_T *dim;
char norm_name[NC_MAX_NAME + 1];
int stat = NC_NOERR;
LOG((2, "%s: ncid 0x%x name %s len %d", __func__, ncid, name,
(int)len));
/* Find our global metadata structure. */
if ((stat = nc4_find_nc_grp_h5(ncid, &nc, &grp, &h5)))
return stat;
assert(h5 && nc && grp);
/* If the file is read-only, return an error. */
if (h5->no_write)
return NC_EPERM;
/* Check some stuff if strict nc3 rules are in effect. */
if (h5->cmode & NC_CLASSIC_MODEL)
{
#ifdef LOOK
/* Only one limited dimenson for strict nc3. */
if (len == NC_UNLIMITED) {
int i;
for(i=0;i<ncindexsize(grp->dim);i++) {
dim = (NC_DIM_INFO_T*)ncindexith(grp->dim,i);
if(dim == NULL) continue;
if (dim->unlimited)
return NC_EUNLIMIT;
}
}
/* Must be in define mode for stict nc3. */
if (!(h5->flags & NC_INDEF))
return NC_ENOTINDEFINE;
#endif
}
/* Make sure this is a valid netcdf name. */
if ((stat = nc4_check_name(name, norm_name)))
return stat;
/* Since unlimited is supported, len >= 0 */
if(len < 0)
return NC_EDIMSIZE;
/* For classic model: dim length has to fit in a 32-bit unsigned
* int, as permitted for 64-bit offset format. */
if (h5->cmode & NC_CLASSIC_MODEL)
if(len > X_UINT_MAX) /* Backward compat */
return NC_EDIMSIZE;
/* Make sure the name is not already in use. */
dim = (NC_DIM_INFO_T*)ncindexlookup(grp->dim,norm_name);
if(dim != NULL)
return NC_ENAMEINUSE;
/* If it's not in define mode, enter define mode. Do this only
* after checking all input data, so we only enter define mode if
* input is good. */
if (!(h5->flags & NC_INDEF))
if ((stat = NCZ_redef(ncid)))
return stat;
/* Add a dimension to the list. The ID must come from the file
* information, since dimids are visible in more than one group. */
if ((stat = nc4_dim_list_add(grp, norm_name, len, -1, &dim)))
return stat;
{
NCZ_DIM_INFO_T* diminfo = NULL;
/* Create struct for NCZ-specific dim info. */
if (!(diminfo = calloc(1, sizeof(NCZ_DIM_INFO_T))))
return NC_ENOMEM;
dim->format_dim_info = diminfo;
diminfo->common.file = h5;
}
/* Pass back the dimid. */
if (idp)
*idp = dim->hdr.id;
return stat;
}
/**
* @internal Find out name and len of a dim. For an unlimited
* dimension, the length is the largest length so far written. If the
* name of lenp pointers are NULL, they will be ignored.
*
* @param ncid File and group ID.
* @param dimid Dimension ID.
* @param name Pointer that gets name of the dimension.
* @param lenp Pointer that gets length of dimension.
*
* @return ::NC_NOERR No error.
* @return ::NC_EBADID Bad ncid.
* @return ::NC_EDIMSIZE Dimension length too large.
* @return ::NC_EBADDIM Dimension not found.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
NCZ_inq_dim(int ncid, int dimid, char *name, size_t *lenp)
{
NC *nc;
NC_FILE_INFO_T *h5;
NC_GRP_INFO_T *grp, *dim_grp;
NC_DIM_INFO_T *dim;
int stat = NC_NOERR;
LOG((2, "%s: ncid 0x%x dimid %d", __func__, ncid, dimid));
/* Find our global metadata structure. */
if ((stat = nc4_find_nc_grp_h5(ncid, &nc, &grp, &h5)))
return stat;
assert(h5 && nc && grp);
/* Find the dimension and its home group. */
if ((stat = nc4_find_dim(grp, dimid, &dim, &dim_grp)))
return stat;
assert(dim);
/* Return the dimension name, if the caller wants it. */
if (name && dim->hdr.name)
strcpy(name, dim->hdr.name);
/* Return the dimension length, if the caller wants it. */
if (lenp)
{
#ifdef LOOK
if (dim->unlimited)
{
/* Since this is an unlimited dimension, go to the file
and see how many records there are. Take the max number
of records from all the vars that share this
dimension. */
*lenp = 0;
if ((stat = ncz_find_dim_len(dim_grp, dimid, &lenp)))
return stat;
}
else
#endif
{
if (dim->too_long)
{
stat = NC_EDIMSIZE;
*lenp = NC_MAX_UINT;
}
else
*lenp = dim->len;
}
}
return stat;
}
/**
* @internal Rename a dimension, for those who like to prevaricate.
*
* @note If we're not in define mode, new name must be of equal or
* less size, if strict nc3 rules are in effect for this file. But we
* don't check this because reproducing the exact classic behavior
* would be too difficult. See github issue #1340.
*
* @param ncid File and group ID.
* @param dimid Dimension ID.
* @param name New dimension name.
*
* @return ::NC_NOERR No error.
* @return ::NC_EBADID Bad ncid.
* @return ::NC_EHDFERR ZARR returned error.
* @return ::NC_ENOMEM Out of memory.
* @return ::NC_EINVAL Name must be provided.
* @return ::NC_ENAMEINUSE Name is already in use in group.
* @return ::NC_EMAXNAME Name is too long.
* @return ::NC_EBADDIM Dimension not found.
* @return ::NC_EBADNAME Name breaks netCDF name rules.
* @return ::NC_EDIMMETA Unable to delete ZARR dataset.
* @author Dennis Heimbigner, Ed Hartnett
*/
int
NCZ_rename_dim(int ncid, int dimid, const char *name)
{
NC_GRP_INFO_T *grp;
NC_DIM_INFO_T *dim;
NC_FILE_INFO_T *h5;
char norm_name[NC_MAX_NAME + 1];
int stat;
/* Note: name is new name */
if (!name)
return NC_EINVAL;
LOG((2, "%s: ncid 0x%x dimid %d name %s", __func__, ncid,
dimid, name));
/* Find info for this file and group, and set pointer to each. */
if ((stat = nc4_find_grp_h5(ncid, &grp, &h5)))
return stat;
assert(h5 && grp);
/* Trying to write to a read-only file? No way, Jose! */
if (h5->no_write)
return NC_EPERM;
/* Make sure this is a valid netcdf name. */
if ((stat = nc4_check_name(name, norm_name)))
return stat;
/* Get the original dim. */
if ((stat = nc4_find_dim(grp, dimid, &dim, NULL)))
return stat;
assert(dim && dim->format_dim_info);
/* Check if new name is in use. */
if (ncindexlookup(grp->dim, norm_name))
return NC_ENAMEINUSE;
/* Give the dimension its new name in metadata. UTF8 normalization
* has been done. */
assert(dim->hdr.name);
free(dim->hdr.name);
if (!(dim->hdr.name = strdup(norm_name)))
return NC_ENOMEM;
LOG((3, "dim is now named %s", dim->hdr.name));
/* rebuild index. */
if (!ncindexrebuild(grp->dim))
return NC_EINTERNAL;
return NC_NOERR;
}